Re: ASAN read-only poisoned by user shadow byte

2019-12-12 Thread 'Alexander Potapenko' via address-sanitizer
On Thu, Dec 12, 2019 at 2:12 PM Andrea Fioraldi wrote: > > Hi Glider, thank you for the quick answer. > I've some doubt, correct me if I'm wrong. > > > No, right now it's not possible. ASan instrumentation doesn't distinguish > > between a read and a write. > > if ((addr >> 3) + offset) __asan_re

Re: ASAN read-only poisoned by user shadow byte

2019-12-12 Thread Andrea Fioraldi
Hi Glider, thank you for the quick answer. I've some doubt, correct me if I'm wrong. > No, right now it's not possible. ASan instrumentation doesn't distinguish between a read and a write. if ((addr >> 3) + offset) __asan_report_load8(addr); This is inserted before loads, a quick implementation

Re: ASAN read-only poisoned by user shadow byte

2019-12-12 Thread Alexander Potapenko
On Thu, Dec 12, 2019 at 1:03 PM Andrea Fioraldi wrote: > > Hi to all, Hi Andrea, > there is a way to poison bytes as read-only in the ASAN runtime? > For instance, I have a type field in a dynamic allocated structure that is > assigned only at creation > and I want to mark it as read-only after

ASAN read-only poisoned by user shadow byte

2019-12-12 Thread Andrea Fioraldi
Hi to all, there is a way to poison bytes as read-only in the ASAN runtime? For instance, I have a type field in a dynamic allocated structure that is assigned only at creation and I want to mark it as read-only after the first assignment to detect type-confusions. I found nothing searching about